Professional Documents
Culture Documents
Preprint 19-003
ASSESSING THE QUALITY OF INCIDENT INVESTIGATIONS AND ITS EFFECT ON SAFETY PERFORMANCE: A STUDY OF THE
GHANAIAN MINING INDUSTRY
ABSTRACT In this paper, the term ‘incidents’ is used more broadly and refers
to undesirable events which affect safety, including those with actual
Incident investigation is of utmost importance in most high-risk consequence and without actual consequence but with a potential to
industries. This study was undertaken to study the content of past do so. Also, the quality of an organisation’s investigation reports is
investigation reports to determine the effectiveness of the incident used as a measure of its safety culture and the effectiveness of its
investigations. The study uses a semi-quantitative method to assess investigation. This is based on the assumption that if an organisation is
the effectiveness of incident investigations in the Ghanaian mining committed to safety, it must reflect in how incidents are investigated. In
industry by evaluating the quality of past investigation reports. The a matured culture, investigators have the courage and management
assessment tool consists of 5 elements with several indicators and support to write about all causal factors that contribute to an incident
rating scales for assessing the quality of an investigation report as a and this must reflect in investigation reports. The nature of the
measure of the effectiveness of the investigation. The method was recommendations must not only be limited to frontline human actions
applied to 304 investigation reports of 3 Ghanaian large-scale gold but ought to be broad addressing technical and organisational related
mines, and the results correlated with incidence rates of the mines to issues as well. The paper first describes the method for assessing the
determine if any relationship existed. The results showed that the quality of investigation reports, followed by results of applying the
mines differ significantly in the quality of their investigation reports, method to investigation reports from 3 mines to indicate the usability
suggesting differences in the effectiveness of their investigations. In and usefulness of the method.
addition, the incidence rates of the mines negatively correlated with
some elements of the assessment tool. In general, the method was METHOD
found useful and revealed areas where improvement is needed.
Adapting a method for assessing the quality of accident
Keyword: investigation reports; incidence rate; causal factors; investigations
corrective measures The assessment method described here was partly adapted from
Jacobsson et al. [2] and Itoh et al. [5], and adding elements specific to
INTRODUCTION incident investigation reports collected from 6 large-scale Ghanaian
The importance of learning from incidents is well documented in mines. The method consists of three components (Table 1, see
the literature. For instance, some scholars regard learning from Appendix), namely,
incidents as important for improving safety [1,2]. Thus, many • elements,
organisations spend considerable resources in identifying lessons from
• indicators of each element and
incidents and subsequently on the process of improvement that follows
• a rating scale for assessing the effectiveness of each
the identification of lessons. It is therefore important for such
element based on the indicators.
organisations to determine the effectiveness of their learning from
incidents processes, in order to identify specific areas where The elements are various components of an investigation report,
improvements are needed, so that interventions can be as specific as and five elements were identified after reading through 450
possible. investigation reports from six mines in Ghana. The elements, in turn,
contain several indicators that should be covered in an investigation
The learning from incidents process models the plan-act-check-do
report. To be able to analyse an incident to determine what the causes
approach of management systems, often consisting of four major
were and recommend measures for preventing future events, a good
stages, namely; data collection and incident analysis, planning
description of the incident with relevant data is a prerequisite.
interventions, implementing interventions and follow-ups and
Therefore, as a minimum, an investigation report should contain
evaluation [3,2]. The investigating and analysis of incidents have been
information on the incident itself, the circumstance surrounding the
regarded as one of the vital stages of the learning process since it is
incident, the causes of the incidents and recommendation of measures
the stage where lessons are identified [3,2,4]. It has also been noted
for preventing future events. Thus, the elements of the assessment
elsewhere in the literature that results of the previous stage of the
tool are not exhaustive but rather a minimum requirement. The first
learning process are important for the next stage [3]. For example,
element general information focuses on characteristics of victims,
ineffective incident analyses could result in the planning of ineffective
incident, equipment involved and those conducting the investigation.
remedial measures. One of the primary aims of incident investigations
The second, incident description consists of a detailed description of
is to identify lessons from previous incidents which when implemented
the incident and the circumstances surrounding it. The third, time-band
can prevent future recurrences or minimise the effects of future
description focuses on details of the sequence of event leading up to
incidents. There is thus the need for organisations to assess the quality
and when the incident was resolved. The fourth and fifth elements
of their investigations periodically since it is one of the means they
respectively concern the quality of causal factors identified from the
adopt to improve safety and the quality of remedial measures to some
incident analysis and the corrective measures proposed following the
extent depends on the effectiveness of the investigations. Although
identified causes.
previous studies [5,2] have assessed the effectiveness of the learning
process, and some aspects of the learning process, such as incidents In order to assess and express the quality of investigation for the
reporting, there is less focus on the effectiveness of incident various elements numerically (based on the content of investigation
investigations. The research described in this paper was therefore reports), a rating scale was adapted from Jacobsson et al. [2]. To
conducted to contribute to this gap, focusing more closely on incident assess the quality of investigation as objective as possible, the
investigation reports. different indicators for each element were described in a semi-
quantitative way on a scale of 1 to 10, divided into four quality levels as
shown in Table 1. In using the scale, the assessor compares the Apart from elements E1 and E2 that showed no significant differences
information in the actual investigation report with the description in the in their mean scores, there were statistically significant differences in
scale, and the level best matching the actual description is selected [2]. the mean scores of the mines for the rest of the elements and the
After assessing each element, a mean value can be computed to overall quality of investigation. Post-hoc testing (Tukey HSD) showed
determine the overall quality of an investigation report as a measure of specific mine pairs with statistically significant differences in their
the effectiveness of the investigation. quality of investigation scores. For instance, for the overall quality of
investigation, mines pair AB and BC had statistically significant
Applying the assessment tool differences in their mean score, whereas, no significant difference was
The assessment tool was applied to 304 investigation reports of 3 found between Mines A and C. The mean scores of A (5.08 ± 1.45)
large-scale gold mines in Ghana. Table 2 shows characteristics of the and C (5.00 ± 0.98) was significantly higher than that of B 4.48 ± 1.05).
mines together with the number of reports. All the mines have a formal
process for managing incidents described in several documents. Table 4. One-way ANOVA of quality of investigation score (significant
Statistical tests were used to determine whether significant differences at p < 0.05).
existed in the quality of investigation among the 3 mines. Element name F(2, 301) Significant
E1. General information 1.02 0.360
Table 2. Attributes of the mines and the number of reports analysed
E2. Incident description 4.75 0.010
(2012-2016).
E3. Time-band description 48.56 0.000
Mine Type Operations Employees Reports E4. Causal factors 13.27 0.000
Part of a major multi- E5. Corrective measures 19.18 0.000
A Open-cast 1872 108
national corporation Overall quality of investigation 7.58 0.001
Part of a major multi-
B Open-cast 4892 104
national corporation Incidence rates as a measure of safety performance
Foreign-owned national Open-cast & The incidence rates of the mines differed from each other. Mine A
C 1603 92
with multiple sites underground had the lowest rate, followed by Mine C, with Mine B having the
highest incidence rate (Table 3). Again, one-way ANOVA (F(2, 177) =
Predictive validity of the assessment tool 29.53, p < 0.05) showed statistically significant difference in the mean
The predictive validity of the assessment tool was examined incidence rates of the mines. Tukey HSD post hoc test showed that
through correlation analysis. The quality of investigation scores for the Mine A’s (0.12 ± 0.11) incidence rate was statistically significantly
mines was correlated with the mine’s incidence rates. First, the lower than that of B (0.22 ± 0.13) and C (0.26 ± 0.05). However, there
monthly incidence rates of the mines were calculated using the was no statistically significant difference between the incidence rates
Australian Standard [6] for workplace injury and disease recording. The of Mines B and C.
incidence rates per one hundred workers were used as a measure of
the safety performance of the mines. The incidents used were the total Predictive validity of the assessment tools
reportable injuries [7]. The 5-year (2012-2016) incidence rate for each The Pearson’s correlation coefficients in Table 5 show that all the
mine was computed. Secondly, the means of the quality of elements of the quality of investigation assessment tool correlated
investigation scores of each of the 5 elements for each mine were negatively with the incidence rate expect element 2, incident
computed. These means were then correlated with the 5-year description, which had a positive correlation. However, it is only the
incidence rate of each mine through Pearson’s correlation. correlations of elements E4 (causal factors) and E5 (corrective
measures) that were found to be statistically significant, with the
RESULTS remaining elements being non-significant. Element E4, causal factor,
Quality of investigation had the highest correlation coefficient of -0.99 at p = 0.015. A
The means of the quality of investigation scores and the 5-year scatterplot (Fig. 1) of the incidence rates and the quality of
incident rates of all 3 mines are shown in Table 3. Some general investigation scores shows that Mine A with the highest quality of
observation can be made from the result. For most of the elements, the investigation score had the lowest incidence rate, whiles Mine B with
quality of investigation was at best “good” for the mines, except for the the highest incidence rate had the lowest quality of investigation score.
elements causal factors and corrective measures, which were “fair” for Table 5. Pearson’s correlation (r) between incidence rates and safety
Mines B, and C. Mine A had the highest scores for most of the culture maturity scores (N = 3, significant at p < 0.05).
elements than the other mines, remarkably for the elements causal Element name Mean SD r p-value
factors and corrective measures. Mine B trails in most of the elements.
E1. General information 6.02 0.1 -0.85 0.153
Across the mines and for the whole cohort, general information (E1)
E2. Incident description 5.66 0.26 0.91 0.090
and incident description (E2) appear to be the reasonably good
E3. Time-band description 4.52 1.03 -0.19 0.809
elements, whereas the important elements causal factors (E4) and
E4. Causal factors 4.39 0.52 -0.99 0.015
corrective measures (E5) seem to be the weakest. Consistently across
E5. Corrective measures 3.57 0.54 -0.97 0.026
the mines, the scores of the causal factors element exceeded that of
Overall quality of investigation 4.85 0.28 -0.79 0.210
corrective measures. This indicates that there were several instances
where causal factors identified were more broadly including aspects of
DISCUSSION
operator errors, workplace conditions and organisational factors.
However, in addressing these causal factors, a narrow view was In this research, a method for assessing the effectiveness of
adopted, focusing more on the proximate causes. incident investigation is presented and subsequently tested in real
case studies. Again, the predictive validity of the method is tested
Table 3. Mean values of quality of investigation scores and incidence
through correlation analysis. The research uses the quality of an
rates.
organisation’s incident investigation reports as a measure of the
E Incidence
Mine E1 E2 E3 E4 E5 effectiveness of its investigation. This is based on the assumption that
average rate
the effectiveness of an organisation’s incident investigation should
A 6.15 5.30 4.46 5.12 4.35 5.08 0.12 reflect in its investigation reports since the reports are the means of
B 5.98 5.81 3.37 3.96 3.17 4.48 0.26 communicating the investigation outcomes to management and
C 5.91 5.91 5.89 4.11 3.25 5.00 0.22 workers. The method was found useful in assessing past investigation
Whole reports, presenting some interesting results. However, a number of
6.02 5.66 4.52 4.39 3.57 4.85 0.19
cohort limitation should be noted in interpreting the results. First is the
subjective nature of the rating scare. Other opinions may exist about
One-way analysis of variance (ANOVA) shows that some of the appropriate rating scale, although the rate was found useful, yielding
quality of investigation elements differentiated between the mines and valuable results. It is therefore recommended that to strengthen the
the mines have different levels of investigating incidents (Table 4). objectivity of the rating scale, future research should focus on testing
2 Copyright © 2019 by SME
SME Annual Meeting
Feb. 24 - 27, 2018, Denver, CO
the validity and reliability of the method proposed here. The second performance of an organisation [8]. In addition, the mine (A) that had
limitation is the sample size; the method was applied to 304 the highest quality of investigation score had the lowest incidence rate,
investigation reports across three mines. Results may therefore not and the mine (B) that had the lowest quality of investigation score had
representing what is happening in the industry. A larger sample size the highest incidence rate. However, further investigation and testing is
across several mines would be of great value to future research in strongly recommended to determine the usefulness and applicability of
determining the effectiveness of investigation across the industry. the method in a wide range of mines and industries.
0.28 CONCLUSIONS
Overall quality of investigation score
APPENDIX
Table 1. Tool for assessing the quality incident investigation.
Element Indicators Quality (effectiveness) levels
Information about victims (e.g. name, age, Covers a few of the most Covers the most Covers all types of the Covers all indicators
General information
gender, work experience, department, job relevant indicators relevant indicators but relevant indicators but that are relevant and
title) date of incident, shift/time of incident, not well organised some indicators not in would add to the
date/time incident was reported, place of details usefulness of the
(E1)
What happened, type/class of incident, Missing relevant information Only the most obvious Almost all the relevant Covers all indicators in
Incident description
activity at the time of incident, damage to and facts on most of the information is indicators are full details, making
mining operation and mine equipment, indicators provided, making it provided, however useful and
condition of workplace, body part injured, hard to undertake some are not covered comprehensive
(E2)
actions taken after the incident, work comprehensive in detail, requiring analysis possible
environment, including the work-site, analysis of the incident additional information
operations, systems, work procedures
The sequence of events leading up to and Covers a few of the Covers the most Covers almost all the All indicators are
including the incident and when the incident indicators, mostly relevant indicators, relevant indicators, but covered in details,
Time-band of
was resolved. This includes information description of the mostly the pre- some not covered in making useful and
description
about past events (long before the incident), occurrence of the accidents accident and accident detail, requiring compressive analysis
(E3)
pre-incident (just before the incident), stage, however not additional information possible
incident (when the accident occurred), post- well organised
incident (immediately after the accident) and
future events (situation to be expected).
A measure of the quality of the causes Only identifies direct causes, Deals with both Covers both direct and All indicators are
determined. It should include various and usually operator errors or operator error and underlying causes and covered and in details,
Causal factors
multiple causal factors, both at the sharp-end some technical failures, a technical failures more begins to include making in-depth
and the blunt-end alike - personal very local view broadly and some aspect of workplace analysis possible
(E4)
administrative controls, human-centred technical changes. including some aspect (with timeline and on a higher
actions and changes in organisational policy, Corrective actions aimed at of procedures, responsible persons). organisational level
believes, norms and culture), identify preventing the same incident training, technical Actions address such as management
(E5)
responsible persons to implement the actions from happening at the same changes, etc., but underlying causes, systems and safety
and specifies the timeframe of the actions. place mostly rather local workplace design, culture. Actions are
view work methods broad, on a site and/or
considerations. company level